browsentic 0.4.6 → 0.4.8

package/dist/cli.js

@@ -28636,12 +28636,9 @@ var CURSOR_PATHS = [
 ];
 var CURSOR2 = `url("data:image/svg+xml,${encodeURIComponent(cursorSvg())}") 14 14, crosshair`;
 var DEFAULT_HINT = "Click the element you mean";
-var picking = false;
-function lensIsUp() {
-  return picking;
-}
+var dismissCurrent = null;
 function pickWithLens({ hint, timeoutMs }) {
-  picking = true;
+  dismissCurrent?.();
   const host = document.createElement("div");
   host.id = HOST_ID;
   host.style.cssText = "all: initial; position: static;";
@@ -28656,6 +28653,8 @@ function pickWithLens({ hint, timeoutMs }) {
   const chip = root.querySelector(".chip");
   let hovered = null;
   return new Promise((resolve) => {
+    const dismiss = () => settle2({ cancelled: true });
+    dismissCurrent = dismiss;
     const timer = setTimeout(() => settle2({ timedOut: true }), timeoutMs);
     const mute = (event) => {
       event.stopPropagation();
@@ -28720,7 +28719,7 @@ function pickWithLens({ hint, timeoutMs }) {
       }
       host.remove();
       cursor.remove();
-      picking = false;
+      if (dismissCurrent === dismiss) dismissCurrent = null;
       resolve(outcome);
     }
   });
@@ -28799,18 +28798,16 @@ function styles() {
 
 // ../lib/actions/page/pick-element.ts
 var MAX_CONTENT = 2e4;
+var PICK_DEFAULT_TIMEOUT_MS = 6e4;
 var pickElement = defineAction({
   name: "page.pickElement",
-  description: "Ask the user to point at an element \u2014 A-Eye. Their cursor becomes a lens, whatever they hover is outlined, and the element they click comes back with its selector, its role and its rendered text. Use it when a target is genuinely ambiguous \u2014 several things share a label, or the user said \u201Cthis one\u201D about something you cannot see \u2014 and pointing is faster than describing. It takes over the page and waits for a person, so never call it to explore, and never call it twice in a row.",
+  description: "Ask the user to point at an element \u2014 A-Eye. Their cursor becomes a lens, whatever they hover is outlined, and the element they click comes back with its selector, its role, its rendered text and a screenshot of it exactly as they saw it. Use it when a target is genuinely ambiguous \u2014 several things share a label, or the user said \u201Cthis one\u201D about something you cannot see \u2014 and pointing is faster than describing. It takes over the page and waits for a person, so never call it to explore; a new call dismisses a pick already waiting.",
   input: external_exports.object({
     hint: external_exports.string().max(120).optional().describe("One line shown over the page saying what to point at, e.g. \u201CPoint at the price you mean\u201D"),
     maxContentLength: external_exports.number().int().positive().max(MAX_CONTENT).default(2e3).describe('Characters of the element\u2019s rendered text to return; past that it is cut and "truncated" comes back true'),
-    timeoutMs: external_exports.number().int().min(5e3).max(3e5).default(6e4).describe("How long to wait for the user to click before giving up")
+    timeoutMs: external_exports.number().int().min(5e3).max(3e5).default(PICK_DEFAULT_TIMEOUT_MS).describe("How long to wait for the user to click before giving up")
   }),
   async execute({ hint, maxContentLength, timeoutMs }) {
-    if (lensIsUp()) {
-      throw new ActionError("A-Eye is already waiting for the user to point at something", "ACTION_FAILED");
-    }
     const outcome = await pickWithLens({ hint, timeoutMs });
     if ("timedOut" in outcome) {
       throw new ActionError(
@@ -28824,12 +28821,18 @@ var pickElement = defineAction({
     const element = outcome.picked;
     const rendered2 = element instanceof HTMLElement ? element.innerText : element.textContent ?? "";
     const content = rendered2.replace(/\n{3,}/g, "\n\n").trim() || accessibleText(element);
+    const rect = element.getBoundingClientRect();
     return {
       element: describeElement(element),
       content: content.slice(0, maxContentLength),
       truncated: content.length > maxContentLength,
       url: location.href,
-      title: document.title
+      title: document.title,
+      capture: {
+        region: { x: rect.x, y: rect.y, w: rect.width, h: rect.height },
+        viewport: { w: window.innerWidth, h: window.innerHeight },
+        dpr: window.devicePixelRatio || 1
+      }
     };
   }
 });
@@ -29895,7 +29898,8 @@ var AGENTS = {
     vendor: "Anthropic",
     bin: "claude",
     install: "npm i -g @anthropic-ai/claude-code",
-    docs: "https://claude.com/claude-code"
+    docs: "https://claude.com/claude-code",
+    models: ["claude-fable-5", "claude-opus-5", "claude-sonnet-5", "claude-haiku-4-5"]
   },
   codex: {
     kind: "codex",
@@ -29903,7 +29907,8 @@ var AGENTS = {
     vendor: "OpenAI",
     bin: "codex",
     install: "npm i -g @openai/codex",
-    docs: "https://developers.openai.com/codex/cli"
+    docs: "https://developers.openai.com/codex/cli",
+    models: ["gpt-5.6-terra", "gpt-5.1-codex-max", "gpt-5.1-codex", "gpt-5.1-codex-mini"]
   },
   antigravity: {
     kind: "antigravity",
@@ -29911,7 +29916,8 @@ var AGENTS = {
     vendor: "Google",
     bin: "agy",
     install: "https://antigravity.google/docs/cli/install",
-    docs: "https://antigravity.google/docs/cli"
+    docs: "https://antigravity.google/docs/cli",
+    models: ["gemini-3-pro", "gemini-3-flash"]
   }
 };
 var AGENT_LIST = AGENT_KINDS.map((kind) => AGENTS[kind]);
@@ -29925,11 +29931,13 @@ var SAVE_SITE_MAP_ACTION = `${RESERVED_PREFIX}saveSiteMap`;
 var START_RECORDING_ACTION = `${RESERVED_PREFIX}startRecording`;
 var STOP_RECORDING_ACTION = `${RESERVED_PREFIX}stopRecording`;
 var READ_SITEMAP_ACTION = `${RESERVED_PREFIX}readSitemap`;
+var FOCUS_SHOT_ACTION = `${RESERVED_PREFIX}focusShot`;
 var RESERVED_ACTIONS = [
   SAVE_SITE_MAP_ACTION,
   START_RECORDING_ACTION,
   STOP_RECORDING_ACTION,
-  READ_SITEMAP_ACTION
+  READ_SITEMAP_ACTION,
+  FOCUS_SHOT_ACTION
 ];
 
 // ../lib/actions/tool-names.ts
@@ -30243,6 +30251,15 @@ var claudeRunner = {
     };
   },
   reader() {
+    let generated = 0;
+    const report = (usage, sink) => {
+      if (!usage) return;
+      generated += usage.output_tokens ?? 0;
+      sink.usage({
+        contextTokens: (usage.input_tokens ?? 0) + (usage.cache_read_input_tokens ?? 0) + (usage.cache_creation_input_tokens ?? 0) + (usage.output_tokens ?? 0),
+        outputTokens: generated
+      });
+    };
     return (line, sink) => {
       const message = parseJsonLine(line);
       if (!message) return;
@@ -30266,10 +30283,14 @@ var claudeRunner = {
           }
           return;
         }
+        case "assistant":
+          if (!message.parent_tool_use_id) report(message.message?.usage, sink);
+          return;
         case "result":
           if (message.is_error) {
             return sink.fail("AGENT_FAILED", message.result || message.subtype || "Claude Code reported an error");
           }
+          if (!generated) report(message.usage, sink);
           return sink.done(message.stop_reason || "end_turn");
       }
     };
@@ -30580,6 +30601,17 @@ var codexRunner = {
             return finish(msg.message, sink);
           case "web_search_begin":
             return sink.tool(randomUUID3(), WEB_TOOL);
+          case "token_count": {
+            const last = msg.info?.last_token_usage ?? msg.info?.total_token_usage;
+            const total = msg.info?.total_token_usage ?? last;
+            if (last) {
+              sink.usage({
+                contextTokens: (last.input_tokens ?? 0) + (last.output_tokens ?? 0),
+                outputTokens: total?.output_tokens ?? 0
+              });
+            }
+            return;
+          }
           case "task_complete":
             return sink.done("end_turn");
           case "error":
@@ -30603,8 +30635,16 @@ var codexRunner = {
           if (kind === "web_search") return sink.tool(item?.id ?? randomUUID3(), WEB_TOOL);
           return;
         }
-        case "turn.completed":
+        case "turn.completed": {
+          const usage = frame.usage;
+          if (usage) {
+            sink.usage({
+              contextTokens: (usage.input_tokens ?? 0) + (usage.output_tokens ?? 0),
+              outputTokens: usage.output_tokens ?? 0
+            });
+          }
           return sink.done("end_turn");
+        }
         case "turn.failed":
           return sink.fail("AGENT_FAILED", frame.error?.message || "Codex could not finish the turn");
         case "error":
@@ -31707,6 +31747,7 @@ function invokeTimeoutFor(action, input2) {
   const declared = input2?.timeoutMs;
   if (typeof declared === "number" && declared > 0) return declared + 1e4;
   if (action === awaitMonitor.name) return AWAIT_DEFAULT_TIMEOUT_MS + 1e4;
+  if (action === pickElement.name) return PICK_DEFAULT_TIMEOUT_MS + 1e4;
   return REQUEST_TIMEOUT_MS;
 }
 
@@ -33455,6 +33496,12 @@ var Server = class extends Protocol {
 // server.ts
 var STATUS_TOOL = toolNameFor(`${RESERVED_PREFIX}status`);
 var SCREENSHOT_TOOL = "page_screenshot";
+var PICK_TOOL = "page_pickElement";
+var FOCUS_SHOT_TOOL = {
+  name: toolNameFor(FOCUS_SHOT_ACTION),
+  description: "Show the screenshot of the element the user pointed at with A-Eye, taken at the instant they picked it. Only answers when the current instruction arrived with a pick attached.",
+  inputSchema: { type: "object", properties: {}, additionalProperties: false }
+};
 var RESOURCES = [
   {
     uri: "browsentic://page/current",
@@ -33568,7 +33615,7 @@ function createMcpServer(bridge, version2, opts = {}) {
           description: "Report whether the Browsentic browser extension is connected, its version, and the active tab. Use this first if a page tool fails.",
           inputSchema: { type: "object", properties: {}, additionalProperties: false }
         },
-        ...opts.agentRun ? [SAVE_SITE_MAP_TOOL] : []
+        ...opts.agentRun ? [SAVE_SITE_MAP_TOOL, FOCUS_SHOT_TOOL] : []
       ]
     };
   });
@@ -33577,6 +33624,8 @@ function createMcpServer(bridge, version2, opts = {}) {
     const action = actionNameFor(params.name);
     const result = await bridge.invoke(action, params.arguments ?? {});
     if (params.name === SCREENSHOT_TOOL) return renderScreenshot(result);
+    if (params.name === FOCUS_SHOT_TOOL.name) return renderFocusShot(result);
+    if (params.name === PICK_TOOL) return renderPick(result, shouldFence(action, policy) ? tag2 : void 0);
     return render(result, shouldFence(action, policy) ? tag2 : void 0);
   });
   server.setRequestHandler(ListResourcesRequestSchema, async () => ({ resources: [...RESOURCES] }));
@@ -33654,6 +33703,38 @@ function renderScreenshot(result) {
     ]
   };
 }
+function renderPick(result, fenceWith) {
+  if (!result.ok) return render(result, fenceWith);
+  const { shot, ...rest } = result.data;
+  if (typeof shot?.dataUrl !== "string") return render(result, fenceWith);
+  const [mimeType, base643] = splitDataUrl(shot.dataUrl);
+  const rendered2 = render({ ok: true, data: rest }, fenceWith);
+  return {
+    content: [
+      ...rendered2.content,
+      { type: "image", data: base643, mimeType },
+      {
+        type: "text",
+        text: `${IMAGE_NOTE} This is the picked element photographed at the instant the user clicked it.`
+      }
+    ]
+  };
+}
+function renderFocusShot(result) {
+  if (!result.ok) return render(result);
+  const dataUrl = result.data?.dataUrl;
+  if (typeof dataUrl !== "string") return render(result);
+  const [mimeType, base643] = splitDataUrl(dataUrl);
+  return {
+    content: [
+      { type: "image", data: base643, mimeType },
+      {
+        type: "text",
+        text: `${IMAGE_NOTE} This is the element the user pointed at with A-Eye, as it stood when they picked it.`
+      }
+    ]
+  };
+}
 function splitDataUrl(dataUrl) {
   const match = /^data:([^;,]+);base64,(.*)$/s.exec(dataUrl);
   return match ? [match[1], match[2]] : ["image/png", ""];
@@ -33679,7 +33760,7 @@ function text2(uri, mimeType, body) {
 // package.json
 var package_default = {
   name: "browsentic",
-  version: "0.4.6",
+  version: "0.4.8",
   description: "Hand your real, logged-in browser to the AI agent you already run. Installs the browser extension, runs the local daemon, and speaks MCP.",
   type: "module",
   license: "MIT",

package/dist/daemon-main.js

@@ -3722,7 +3722,7 @@ var import_websocket = __toESM(require_websocket(), 1);
 var import_websocket_server = __toESM(require_websocket_server(), 1);
 
 // ../lib/actions/protocol.ts
-var SOCKET_PROTOCOL_VERSION = 14;
+var SOCKET_PROTOCOL_VERSION = 16;
 var EXTERNAL_RUN_ID = "external";
 var DAEMON_PORTS = [8765, 8766, 8767];
 var EXTENSION_REQUEST_FRAMES = [
@@ -3740,6 +3740,7 @@ var EXTENSION_REQUEST_FRAMES = [
   "analyzeRecording",
   "agentState",
   "setAgent",
+  "setAgentModel",
   "grantAgent",
   "listSkills",
   "guardrails",
@@ -20228,12 +20229,9 @@ var CURSOR_PATHS = [
 ];
 var CURSOR2 = `url("data:image/svg+xml,${encodeURIComponent(cursorSvg())}") 14 14, crosshair`;
 var DEFAULT_HINT = "Click the element you mean";
-var picking = false;
-function lensIsUp() {
-  return picking;
-}
+var dismissCurrent = null;
 function pickWithLens({ hint, timeoutMs }) {
-  picking = true;
+  dismissCurrent?.();
   const host = document.createElement("div");
   host.id = HOST_ID;
   host.style.cssText = "all: initial; position: static;";
@@ -20248,6 +20246,8 @@ function pickWithLens({ hint, timeoutMs }) {
   const chip = root.querySelector(".chip");
   let hovered = null;
   return new Promise((resolve4) => {
+    const dismiss = () => settle2({ cancelled: true });
+    dismissCurrent = dismiss;
     const timer = setTimeout(() => settle2({ timedOut: true }), timeoutMs);
     const mute = (event) => {
       event.stopPropagation();
@@ -20312,7 +20312,7 @@ function pickWithLens({ hint, timeoutMs }) {
       }
       host.remove();
       cursor.remove();
-      picking = false;
+      if (dismissCurrent === dismiss) dismissCurrent = null;
       resolve4(outcome);
     }
   });
@@ -20391,18 +20391,16 @@ function styles() {
 
 // ../lib/actions/page/pick-element.ts
 var MAX_CONTENT = 2e4;
+var PICK_DEFAULT_TIMEOUT_MS = 6e4;
 var pickElement = defineAction({
   name: "page.pickElement",
-  description: "Ask the user to point at an element \u2014 A-Eye. Their cursor becomes a lens, whatever they hover is outlined, and the element they click comes back with its selector, its role and its rendered text. Use it when a target is genuinely ambiguous \u2014 several things share a label, or the user said \u201Cthis one\u201D about something you cannot see \u2014 and pointing is faster than describing. It takes over the page and waits for a person, so never call it to explore, and never call it twice in a row.",
+  description: "Ask the user to point at an element \u2014 A-Eye. Their cursor becomes a lens, whatever they hover is outlined, and the element they click comes back with its selector, its role, its rendered text and a screenshot of it exactly as they saw it. Use it when a target is genuinely ambiguous \u2014 several things share a label, or the user said \u201Cthis one\u201D about something you cannot see \u2014 and pointing is faster than describing. It takes over the page and waits for a person, so never call it to explore; a new call dismisses a pick already waiting.",
   input: external_exports.object({
     hint: external_exports.string().max(120).optional().describe("One line shown over the page saying what to point at, e.g. \u201CPoint at the price you mean\u201D"),
     maxContentLength: external_exports.number().int().positive().max(MAX_CONTENT).default(2e3).describe('Characters of the element\u2019s rendered text to return; past that it is cut and "truncated" comes back true'),
-    timeoutMs: external_exports.number().int().min(5e3).max(3e5).default(6e4).describe("How long to wait for the user to click before giving up")
+    timeoutMs: external_exports.number().int().min(5e3).max(3e5).default(PICK_DEFAULT_TIMEOUT_MS).describe("How long to wait for the user to click before giving up")
   }),
   async execute({ hint, maxContentLength, timeoutMs }) {
-    if (lensIsUp()) {
-      throw new ActionError("A-Eye is already waiting for the user to point at something", "ACTION_FAILED");
-    }
     const outcome = await pickWithLens({ hint, timeoutMs });
     if ("timedOut" in outcome) {
       throw new ActionError(
@@ -20416,12 +20414,18 @@ var pickElement = defineAction({
     const element = outcome.picked;
     const rendered2 = element instanceof HTMLElement ? element.innerText : element.textContent ?? "";
     const content = rendered2.replace(/\n{3,}/g, "\n\n").trim() || accessibleText(element);
+    const rect = element.getBoundingClientRect();
     return {
       element: describeElement(element),
       content: content.slice(0, maxContentLength),
       truncated: content.length > maxContentLength,
       url: location.href,
-      title: document.title
+      title: document.title,
+      capture: {
+        region: { x: rect.x, y: rect.y, w: rect.width, h: rect.height },
+        viewport: { w: window.innerWidth, h: window.innerHeight },
+        dpr: window.devicePixelRatio || 1
+      }
     };
   }
 });
@@ -21489,6 +21493,7 @@ var SAVE_SITE_MAP_ACTION = `${RESERVED_PREFIX}saveSiteMap`;
 var START_RECORDING_ACTION = `${RESERVED_PREFIX}startRecording`;
 var STOP_RECORDING_ACTION = `${RESERVED_PREFIX}stopRecording`;
 var READ_SITEMAP_ACTION = `${RESERVED_PREFIX}readSitemap`;
+var FOCUS_SHOT_ACTION = `${RESERVED_PREFIX}focusShot`;
 
 // ../lib/agents/catalog.ts
 var AGENT_KINDS = ["claude", "codex", "antigravity"];
@@ -21500,7 +21505,8 @@ var AGENTS = {
     vendor: "Anthropic",
     bin: "claude",
     install: "npm i -g @anthropic-ai/claude-code",
-    docs: "https://claude.com/claude-code"
+    docs: "https://claude.com/claude-code",
+    models: ["claude-fable-5", "claude-opus-5", "claude-sonnet-5", "claude-haiku-4-5"]
   },
   codex: {
     kind: "codex",
@@ -21508,7 +21514,8 @@ var AGENTS = {
     vendor: "OpenAI",
     bin: "codex",
     install: "npm i -g @openai/codex",
-    docs: "https://developers.openai.com/codex/cli"
+    docs: "https://developers.openai.com/codex/cli",
+    models: ["gpt-5.6-terra", "gpt-5.1-codex-max", "gpt-5.1-codex", "gpt-5.1-codex-mini"]
   },
   antigravity: {
     kind: "antigravity",
@@ -21516,7 +21523,8 @@ var AGENTS = {
     vendor: "Google",
     bin: "agy",
     install: "https://antigravity.google/docs/cli/install",
-    docs: "https://antigravity.google/docs/cli"
+    docs: "https://antigravity.google/docs/cli",
+    models: ["gemini-3-pro", "gemini-3-flash"]
   }
 };
 var AGENT_LIST = AGENT_KINDS.map((kind) => AGENTS[kind]);
@@ -21642,6 +21650,21 @@ function writeActiveAgent(kind) {
   const stored = readStored();
   write({ ...stored, agent: kind });
 }
+function writeAgentModel(kind, model) {
+  const stored = readStored();
+  const agents = { ...stored.agents ?? {} };
+  const scoped = { ...agents[kind] ?? {} };
+  const value = text(model);
+  if (value) scoped.model = value;
+  else delete scoped.model;
+  if (Object.keys(scoped).length) agents[kind] = scoped;
+  else delete agents[kind];
+  const next = { ...stored };
+  if (Object.keys(agents).length) next.agents = agents;
+  else delete next.agents;
+  if (kind === "claude") delete next.model;
+  write(next);
+}
 function writeGuardrailSetting(setting, value) {
   const stored = readStored();
   const guardrails = { ...stored.guardrails ?? {} };
@@ -23346,6 +23369,11 @@ function revokeSessions(predicate) {
 // agent/service.ts
 import { randomUUID as randomUUID6 } from "crypto";
 
+// ../lib/actions/tool-names.ts
+function toolNameFor(actionName) {
+  return actionName.replaceAll(".", "_");
+}
+
 // ../lib/skills/scrub.ts
 var CONTROL_CHARS = new RegExp(
   "[\\u0000-\\u0008\\u000b-\\u001f\\u007f-\\u009f\\u200b-\\u200f\\u202a-\\u202e\\u2066-\\u2069\\ufeff]",
@@ -23615,6 +23643,15 @@ var claudeRunner = {
     };
   },
   reader() {
+    let generated = 0;
+    const report = (usage, sink) => {
+      if (!usage) return;
+      generated += usage.output_tokens ?? 0;
+      sink.usage({
+        contextTokens: (usage.input_tokens ?? 0) + (usage.cache_read_input_tokens ?? 0) + (usage.cache_creation_input_tokens ?? 0) + (usage.output_tokens ?? 0),
+        outputTokens: generated
+      });
+    };
     return (line, sink) => {
       const message = parseJsonLine(line);
       if (!message) return;
@@ -23638,10 +23675,14 @@ var claudeRunner = {
           }
           return;
         }
+        case "assistant":
+          if (!message.parent_tool_use_id) report(message.message?.usage, sink);
+          return;
         case "result":
           if (message.is_error) {
             return sink.fail("AGENT_FAILED", message.result || message.subtype || "Claude Code reported an error");
           }
+          if (!generated) report(message.usage, sink);
           return sink.done(message.stop_reason || "end_turn");
       }
     };
@@ -23952,6 +23993,17 @@ var codexRunner = {
             return finish(msg.message, sink);
           case "web_search_begin":
             return sink.tool(randomUUID4(), WEB_TOOL);
+          case "token_count": {
+            const last = msg.info?.last_token_usage ?? msg.info?.total_token_usage;
+            const total = msg.info?.total_token_usage ?? last;
+            if (last) {
+              sink.usage({
+                contextTokens: (last.input_tokens ?? 0) + (last.output_tokens ?? 0),
+                outputTokens: total?.output_tokens ?? 0
+              });
+            }
+            return;
+          }
           case "task_complete":
             return sink.done("end_turn");
           case "error":
@@ -23975,8 +24027,16 @@ var codexRunner = {
           if (kind === "web_search") return sink.tool(item?.id ?? randomUUID4(), WEB_TOOL);
           return;
         }
-        case "turn.completed":
+        case "turn.completed": {
+          const usage = frame.usage;
+          if (usage) {
+            sink.usage({
+              contextTokens: (usage.input_tokens ?? 0) + (usage.output_tokens ?? 0),
+              outputTokens: usage.output_tokens ?? 0
+            });
+          }
           return sink.done("end_turn");
+        }
         case "turn.failed":
           return sink.fail("AGENT_FAILED", frame.error?.message || "Codex could not finish the turn");
         case "error":
@@ -24087,6 +24147,7 @@ async function probe2(runner, settings) {
     return {
       kind: runner.kind,
       bin: settings.bin,
+      model: settings.model,
       ready: false,
       problem: {
         code: "AGENT_MISSING",
@@ -24099,6 +24160,7 @@ async function probe2(runner, settings) {
     return {
       kind: runner.kind,
       bin: settings.bin,
+      model: settings.model,
       ready: false,
       problem: {
         code: "AGENT_UNUSABLE",
@@ -24108,7 +24170,14 @@ async function probe2(runner, settings) {
     };
   }
   const problem = await runner.check?.(settings) ?? null;
-  return { kind: runner.kind, bin: settings.bin, ready: !problem, version: found.detail, problem: problem ?? void 0 };
+  return {
+    kind: runner.kind,
+    bin: settings.bin,
+    model: settings.model,
+    ready: !problem,
+    version: found.detail,
+    problem: problem ?? void 0
+  };
 }
 function version2(bin, args) {
   return new Promise((resolve4) => {
@@ -24614,6 +24683,7 @@ function runStream(runner, context, signal, emit) {
       session: (id) => {
         if (id) sessionId = id;
       },
+      usage: (usage) => emit({ kind: "usage", usage }),
       done: (stopReason) => settle2(() => {
         flush();
         resolve4({ stopReason, sessionId });
@@ -25287,6 +25357,7 @@ function isPrivateAddress(address) {
 }
 
 // agent/service.ts
+var FOCUS_SHOT_TOOL_NAME = toolNameFor(FOCUS_SHOT_ACTION);
 var AgentSession = class {
   constructor(deps) {
     this.deps = deps;
@@ -25333,6 +25404,16 @@ var AgentSession = class {
       emit({ kind: "toolResult", toolId, ok: result2.ok, summary: summarize(input2, result2) });
       return result2;
     }
+    if (action === FOCUS_SHOT_ACTION) {
+      const result2 = run.focusShot ? success({ dataUrl: run.focusShot }) : failure("NO_FOCUS_SHOT", "No A-Eye pick came with this instruction, so nothing was photographed.");
+      emit({
+        kind: "toolResult",
+        toolId,
+        ok: result2.ok,
+        summary: result2.ok ? "the picked element, as photographed" : "no pick attached"
+      });
+      return result2;
+    }
     if (run.map) {
       const gate = gateMappingInvoke(run.map, action, input2);
       if (!gate.allow) {
@@ -25442,7 +25523,8 @@ var AgentSession = class {
       site: siteOf(context?.url),
       abort: new AbortController(),
       pending: /* @__PURE__ */ new Map(),
-      ownedTabIds: []
+      ownedTabIds: [],
+      focusShot: context?.focus?.shot
     };
     this.runs.set(runId, run);
     const runner = activeRunner(await agentState(config2));
@@ -25648,6 +25730,11 @@ function focusBlock(focus) {
   ];
   if (focus.label) lines.push(`- Label: ${flatten2(focus.label)}`);
   lines.push(`- On: ${flatten2(focus.title)} \u2014 ${flatten2(focus.url)}`);
+  if (focus.shot) {
+    lines.push(
+      `- A screenshot of it, taken at the instant they picked it, is attached \u2014 call \`${FOCUS_SHOT_TOOL_NAME}\` once to see it.`
+    );
+  }
   const content = focus.content.slice(0, MAX_FOCUS_CONTENT);
   const cut = focus.truncated || content.length < focus.content.length;
   lines.push(
@@ -26297,6 +26384,7 @@ function timeoutFor(action, input2) {
   if (action === typeText.name) return typingDurationMs(input2) + DEFAULT_TIMEOUT_MS2;
   if (action === startMonitor.name) return DEFAULT_TIMEOUT_MS2;
   if (action === awaitMonitor.name) return (declaredTimeout(input2) ?? AWAIT_DEFAULT_TIMEOUT_MS) + 5e3;
+  if (action === pickElement.name) return (declaredTimeout(input2) ?? PICK_DEFAULT_TIMEOUT_MS) + 5e3;
   const declared = declaredTimeout(input2);
   return declared != null ? declared + 5e3 : DEFAULT_TIMEOUT_MS2;
 }
@@ -26525,7 +26613,7 @@ async function startDaemon({ version: version3, idleExit = true }) {
           );
           return;
         }
-        if (request.t === "agentState" || request.t === "setAgent" || request.t === "grantAgent") {
+        if (request.t === "agentState" || request.t === "setAgent" || request.t === "setAgentModel" || request.t === "grantAgent") {
           void settleAgent(request).then((state) => {
             source.send({ t: "agentInfo", id: request.id, result: state });
             if (request.t === "setAgent") pushSkillCatalog(source);
@@ -26584,6 +26672,11 @@ async function startDaemon({ version: version3, idleExit = true }) {
       agent?.handle({ t: "reset" });
       log(`agent set to ${AGENTS[request.agent].label}`);
     }
+    if (request.t === "setAgentModel") {
+      const model = typeof request.model === "string" ? request.model : null;
+      writeAgentModel(request.agent, model);
+      log(`${AGENTS[request.agent].label} model set to ${model?.trim() || "the default"}`);
+    }
     if (request.t === "grantAgent") await grantRunner(request.agent);
     const refresh = request.t !== "agentState" || request.refresh === true;
     return success(await agentState(readAgentConfig(), { refresh }));
@@ -26874,7 +26967,7 @@ X-Browsentic-Reason: ${reason}\r
 // package.json
 var package_default = {
   name: "browsentic",
-  version: "0.4.6",
+  version: "0.4.8",
   description: "Hand your real, logged-in browser to the AI agent you already run. Installs the browser extension, runs the local daemon, and speaks MCP.",
   type: "module",
   license: "MIT",